Monopolist Pricing with Dynamic Demand and Production Cost

Author(s): Shlomo Kalish

Source: Marketing Science, Vol. 2, No. 2 (Spring, 1983), pp. 135-159
Published by: INFORMS
Stable URL: http://www.jstor.org/stable/184114
Accessed: 21/08/2010 21:42
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=informs.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Marketing Science.

http://www.jstor.org

MONOPOLIST PRICING WITH DYNAMIC

DEMAND AND PRODUCTION COST*
SHLOMO KALISHt
This paper deals with pricing of a new product over time by a monopolist who
maximizes the discounted profit stream. The interdependency of cost and demand on
cumulative production makes the problem inherently dynamic. Cost is assumed to be
declining with cumulative production (learning curve effect), while demand is a
function of price and cumulative sales, representing word-of-mouth and saturation
effects.
The paper addresses this problem in a general framework that includes several
previous results as special cases, and provides new insights in other situations. While
the learning curve and word-of-mouth effect cause prices to be lower than the price
that maximizes immediate revenues, the saturation factor has the opposite effect. The
price path over time is affected by these factors and the interest rate. We characterize
the price path under several different situations and interpret the results for policy
guidelines.
(Pricing; Dynamic Pricing; Learning Curve; Diffusion of Innovations)

1. Introduction
Considerations of the effects of current price and sales on future sales and
production cost did not receive much attention in the pricing literature until
recently. In the last several years, there has been increasing interest in this
subject. This paper addresses the monopolist pricing problem in a general
framework that includes several previous results as special cases, and provides
new insight in other situations.
While most pricing research focuses on steady-state situations in which the
pricing rule is based on marginal cost equalling marginal revenues, it is the
*Received January 1982. This paper has been with the author for 2 revisions.
tGraduate School of Management, University of Rochester, Rochester, New York 14627.
The author would like to thank his colleagues, Anne Coughlan and Subrata Sen, and two
anonymous referees for their helpful comments.
135
MARKETING SCIENCE
Vol. 2. No. 2, Spring 1983
Printed in U.S.A.

0732-2399/83/0202/0135$01.25

Copyright ? 1983, The Institute of Management Sciences

136

SHLOMO KALISH

existence of factors that cause sales and costs to vary over time and makes the
pricing rule over the product life cycle different. These factors include: (1)
'learning curve' cost decline on the production side, the process by which unit
cost declines as more and more experience is gained. This phenomenon has
been studied in the literature (Alchian (1959), Hirsch (1952), Arrow (1962),
Rosen (1972), and Spence (1981)), although it gained popular recognition
more recently. (See, for example, Abell and Hammond (1979), Henderson
(1980), Porter (1980), and Conley (1981).) (2) 'Diffusion' effect on the demand
side. By diffusion we refer to a host of phenomena that causes the likelihood
of purchase to increase, as a result of a higher market penetration. These
phenomena include increased awareness through word of mouth (Rogers and
Shoemaker (1971)), and product self-advertising, reduction in uncertainty and
improved reputation of the product (Jeuland (1981b), Kalish (1983), Schmalensee (1982)), and other factors (Bass (1969), (1980)). (3) Saturation effects for
durable goods, i.e., the fact that for durable goods as more units are sold the
remaining unfulfilled demand decreases. (4) Uncertainty about future technological developments and competitive reaction to follow such developments in
the future (Teng and Thompson (1980), Eliashberg and Jeuland (1982), Clarke
and Doland (1982), and Bass and Rao (1982)).
We restrict the following work to deal with the first three phenomena, i.e.,
the case of a monopolist where there is no direct competition (e.g. patent
protection). We also assume that consumers do not change their behavior by
developing price expectations. Thus, we look at the problem of maximizing
the present value of the profit stream, given that marginal cost of production
decreases with cumulative output, and that the demand for the product
depends on current price, and cumulative sales, representing the diffusion and
saturation effects.
Since the three factors mentioned earlier cause future profits to depend on
past sales, the monopolist has to trade off between current profits and future
profits. The learning curve cost decline alone creates an incentive to reduce
price and produce more, so that future costs will be lower. Similarly, a positive
diffusion effect creates an incentive to price lower than the immediate profit
maximizing price, and accelerate the diffusion of the product that way. On the
other hand, saturation in the case of durable goods creates the opposite effect.
Here, the monopolist wants to price higher, and reduce price as more is sold,
thereby charging a higher price for those who are willing to pay more for early
adoption. The interactions of the above factors, combined with the level of
discount rate-that weighs current and future profits-creates a host of
possible pricing patterns.
There are two types of questions that we try to answer. First, how will price
at each point in time be, relative to the price that will maximize the immediate
profits at that point? Second, what would the price pattern over time look
like? Since cost data are seldom available publicly, it is difficult for the
outsider to compare actual prices to myopic prices. However, price patterns
over time are observable, and we will compare our results with pricing
patterns observed in industry.

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

137

The paper is structured as follows: In ?2 we formulate the general model,

characterize the solution, and discuss the two questions of interest in light of
the different factors. In ?3 we analyze subclasses of the general formulation.
We first look at the effects of learning curve alone, where sales are a function
of price only, in order to isolate this factor from demand complications. Then
we analyze the subclass where sales can be written as a product of the effect of
price and the diffusion/saturation factor. This subclass includes many of the
previously published studies (Robinson and Lakhani (1975), Dolan and Jeuland (1981), Jeuland and Dolan (1982), Spremann (1981), Teng and Thompson (1980)). Finally we look at another subclass, that is more suitable for
durable goods. This is followed with a summary of the results and conclusions
in ?4.
2.

Model Formulation

We begin the analysis by specifying the model with only a few robust
assumptions. This allows us to examine the most general and robust conclusions. Subsequently, we look at more specific cases and obtain stronger results.
Let x(t) be the experience at time t, i.e., the total volume produced (sold) by
that time. The sales incurred from t to t + 1 is clearly x(t + 1)- x(t). We will
use a continuous approximation, by looking at very small time periods; the
sales rate then is the time derivative of x(t): s(t) = dx(t)/dt =-x(t). (We
adopt the dot notation for the derivative with respect to time.)
Sales is a function of price, p(t), as well as the cumulative sales x(t). We
write then: s(t) = x(t) = f(x(t), p(t)). The unit cost, or marginal cost, is a
function of experience x(t). Write c = c(x(t)).
We assume that the increasing price will result in reduced sales, i.e.,
af(x, p)/lp < 0, and that there are learning economies, i.e. dc(x)/dx < 0.
The effect of the experience on sales, af(x, p)lax, is not uniform however.
At introduction, for a 'good' product, it will have a positive effect due to word
of mouth, improved reputation, etc. This will be contrasted, in the case of
durable goods, with the negative effect on demand, since each additional unit
sold removes the customer from the market, thus reducing demand. We will
study these situations in what follows.
The monopolist's problem is to determine a pricing strategy, p(t), that will
maximize the discounted profits over the planning horizon. Let r be the
discount rate, and T the end of the planning horizon; then the sum of the
discounted profits is:
qr=

e-rt. (p(t) - c(x(t)))

s(t)dt

'Modeling marginal cost as a function of experience alone is common. See e.g. Spence (1981),
Dolan and Jeuland (1981), Jeuland and Dolan (1982), Bass (1980) and others. Haley (1981) and
Clarke et al. (1982) model it as a function of experience and rate of production. For a discussion
of these issues see Alchian (1959).

138

SHLOMO KALISH

and the monopolist problem is:

-

Max
e(t)
s.t.

e-rt'(p(t)x(t) =f(x, p);

c(x(t))) x*(t)dt
(1)
x(O) = x.

This is a dynamic optimization problem, and we use the maximum principle

to characterize the optimal policy.2 We introduce the constraint into the
objective function, by multiplying them with the shadow prices, e-' .

X(t), to

form the Hamiltonian:

A) =

H(p,x,

e-rt.

[p - c(x) +

f(x, p).

The maximum principle states that the optimal solution, p*(t), to system 1, has
to maximize the Hamiltonian, H, at each instant t, with the sales x* and the
shadow A* follows the differential equations:
* = f(x*,p*);

*== r*
rX*

= xo,

x*(0)

(2a)

aH*
ax

af(x*,p*)
rf= - (*
(P*(p c(x*) + X*) + c'(x*) f(x*, p*);

A*(T) = 0.

(2b)

(In what follows we will eliminate the function arguments where there is no
confusion, in order to improve clarity.) We assume that the optimal solution
exists at every time t,3 and therefore the derivative of the Hamiltonian with
respect top must vanish on the optimal path p*:

-H-0>p*=
c* - X* ^ -c-X-

ijaH
Op

^
af*t/ap

(2c)

(2c) can be rewritten as:

* =

(c* - *)

(3)

2See Arrow and Kurz (1970) for a review of dynamic optimization techniques with special focus
on economic problems.
3See Appendix for a discussion of this assumption.

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

139

where rqis the elasticity of demand (7 = -(dx/dp). (p/x)). Note that equation (3) resembles the classical static monopolist's pricing rule, that price
markup over marginal cost depends on demand elasticities as above, except
that here we modify marginal costs by subtracting X*, the shadow price.
Recall that the shadow price, X(t), is the net benefit of having the constraint
relaxed by one unit. In our context, it is the dollar value (at time t) of having
one more unit produced. It is exactly because of the experience curve and the
influence of experience on demand that producing one more unit now will
affect future benefits. Thus, for example, if there is only learning in production, producing one more unit now reduces future costs, so X(t) is positive and
marginal cost has to be modified to account for that, causing price to be
lowered. (Alternatively, equation (2c) above says that current marginal revenues plus future marginal benefit must equal marginal cost.4) So we see that
the sign of the shadow price, X(t), will determine whether the price will be
lower or higher than a myopic monopolist facing the same conditions. If
X(t) > 0, i.e. there are future benefits for overproducing, then price will be
lower and vice versa. To see how the shadow price evolves over time, we
substitute (2c) into (2b), and rearrange terms to get:
X=rA+c_X= rX +

fxp ;
.

X(T)=O
(T) =

which has the solution:

xJ(t (=

x -c )er

d.

By examining the equation for X(t), we see that if the 'diffusion' effect on the
demand side is uniformly positive (i.e. fx > 0), then the integrand is uniformly
positive, and so is X(t). Moreover, for the case r = 0, X(t) is monotonically
decreasing. This is consistent with intuition, since in this case there is an
incentive to sacrifice profits now in order to benefit later. On the other
extreme, if c(x) = 0, i.e. no cost decline, and negative diffusion effect on
demand (i.e. fx < 0), then X(t) < 0, and the price will be higher than the
myopic monopolist's price.
The second-order condition for a maximum which we use in what follows
is:

2H < 0

di'(p*

- c* - X*) + 2

dPO

0.

(4)

4See also Simon (1982) for an interpretation of shadow prices in the context of strategic pricing.

140

SHLOMO KALISH

Substituting p* from (2c) we get:

aP2

ap

|P.

f
=[2-

f*2

>1 0

<?0[2-

(5)

*2

(where f = (af(x, p)/ap), fpp= a2(x, p)/ap2, etc. In what follows we will
eliminate the * notation, and all equations refer to the optimal solution, unless
otherwise stated.)
A second question of interest is whether price increases or decreases over
time. In order to analyze that, we take the time derivative of the optimal price
given by (2c):5
c'(x).-A

Substituting (2b) for A, and rearranging terms, we get:

4p2-

fr
p2

1=-rA-2fXhf?f2f2
fp

(6)
p2

Recall from (5) that on the optimal path the term in brackets on the left-hand
side of the equation is positive. Therefore, the sign of the time derivative of
price is the same sign as the right-hand side. It is evident that for the case of
no discount (i.e. r = 0), the first term vanishes, and the dependence of price
derivative on the cost is only implicit through the terminal condition p(T)
= (q/(rt - 1)) c(x(T)) (from (3)), since in this case (2a) and (6) characterize
the optimal solution completely. Thus, in the zero interest case, the price
derivative depends on demand specification f. Hence, for demand functions
where the sign of fx is the sign of fx uniformly, price will be monotonic
regardless of the cost function. In other words, if demand increases with
penetration (fx > 0), and this increase is stronger for higher price (fp > 0),
then price increases, and vice versa. We will see several examples later on.
3.

Subclasses of the General Formulation

While the general formulation discussed in the last section is useful to gain
insight into the factors affecting the optimal price, we can obtain stronger
results for specific cases. First, we consider the case where demand exhibits no
'learning' or saturation, i.e. sales depend on price alone. This case pinpoints
the effects of the learning curve alone. Next, we consider the case where sales
depend on experience in a specific form, i.e. multiplicative form. Finally, we
consider several models that are not included in the above characterizations.
5See Appendix for a discussion of conditions where this can be 'safely' done.

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

141

3.1. Static Demand with Cost Decline

In this section we investigate the effects of learning curve cost decline alone,
where sales are a function of current price alone. This situation characterizes
nondurable products, where word of mouth is not important, e.g. established
products.
In this case (2a)-(2c) reduces to:
(a)

= f(p);

(b)

x(O) = xo,

= rX + c'(x) f,
(7)

1 (c(x)-

(c)Mp=

X),

and the time derivative of the optimal price:

P(2

ff

-r

(8)

Clearly, for r = 0, the problem reduces to a static problem, where the whole
planning period is looked at as one period, and price is set for the whole
period, so that marginal revenues equal marginal cost at the end of the period.
Note that X(t) in this case is: X(t) = c(x(t)) - c(x(T)). This means that if one
or more units are sold at time t, then the benefit is that all subsequent units
will be less expensive. The difference amounts to selling one more unit at the
end, and one less at time t, giving rise to the above benefit.
For r > 0, however, the results change. It is characterized as follows.
If demand is a function of price alone, and costs decrease with
then:
experience,
1. Optimalprice decreases monotonicallyover time for r > 0.
2. Total quantitysold and averageprice are greater and lower respectivelythan
for a myopic monopolist.
3. The present value of the shadow price, e-r *X(t), decreases monotonically
over time.
THEOREM 1.

PROOF. See Appendix.6

This result states that the effect of learning economies is to reduce price and
produce more, in order to benefit by lowering future production cost. Although early production is discounted relatively to the price charged by a
6A similar result was obtained independently by Clarke et al. (1982) for a somewhat more
general cost function. They have also investigated a situation where the fixed cost declines with
experience, but not the marginal unit cost. Interestingly enough, the results reverse in this case,
and price monotonically increases over time. This, however, is a less realistic situation in my
opinion. Previously, special cases of Theorem I were obtained by Jeuland and Dolan (1982), for a
particular cost decline, and Spence (1981) for the no interest case. It is easy to verify that the
theorem actually applies to a wider class of problems, where there is an exogenous shift in
demand, i.e. x = k(t) .f(p). This generalizes the work of Bass and Bultez (1982). See Appendix.

142

SHLOMO KALISH

cumulative

-1 (T)

72 (T)

profit
,",

unit cost
unit
&
unit price
price pUNIT
~~&

Il

5_

rP2

PRICE
UNIT COST

X2 (T) X1 (T)

r=
r> 0

Case 1:
Case 2

cumulative

units sold

Optimal Price and Profits for r = 0 and r > 0.

FIGURE 1.

myopic monopolist, this will always result in a monotonically decreasing price

over time (for r > 0). Moreover, for the case r = 0, we say that the shadow
price, X(t), is monotonically decreasing, resulting in a monotonically decreasing price discount relative to a myopic monopolist facing the same conditions,
and a monotonically increasing profit margin. While this will not be generally
true for the positive interest rate, it may hold for low discount rates.
EXAMPLE 1. Consider a demand function x = k edp.
Then we get:
p* =
For r = O, p* =

X+
x(T) xOx=

/d + c - X,

/d + c(x(T)),

xo+

p* =-

r < O.

where

ke

d-(/d+c(x(T)))

dt=

x + T -k

e-+d((T))

which is an algebraic equation for x(T). Therefore, the firm may actually lose
money at the beginning, depending on whether c(xO) - c(x(T)) > I /d or not.
In the r > 0 case, price starts higher, and decreases monotonically (see
Figure 1).
3.2.

Multiplicative

Separable

Functions

We turn next to investigate the case where the interaction between price and
cumulative sales takes a specific form, i.e. sales, s(x, p) = f(x) g(p). This
subclass of models is interesting for two reasons: (1) It is a naturally simple
way to model the interaction between price and experience. (2) Several

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

143

previous studies that examined pricing over the life cycle have used such
functional forms. For example, Dolan and Jeuland (1981) have used x =
[N - x][a + bx]. e-dP. A similar model was used by Robinson and Lakhani
(1975).
The main implication of this class of formulation is that demand elasticity,
i.e. the percentage response in sales to a percentage change in price, is
independent of experience, x(t). This assumption is particularly bothersome
for durable goods, since in this case the potential population actually changes
as a result of past sales.7 However, for repeat purchase goods this class of
models might be a useful approximation where the experience, x, may stand
for a measure of reputation, for example.
In this case, the system of equation that characterizes the optimal solution,
(2), reduces the following:
(a) x =f(x)

(b)

x(O) = xO,

g(p);

= rX + c'(x). fg

+ f

(c)p = c -

X;(T)

= 0,

(9)

-g

and the derivative of price:

2-g

=-rx-

(10)

3.2.1. Separable Demand, No Discounting. Consider first the case of zero

discount rate. Although this case may sound unrealistic, it is useful to look at
it, since the results are interesting and it serves as an approximation to a low
discount rate case. The following result holds for this case:
THEOREM2.

For the separable demand and zero discount rate, optimal

is
monopolistprice increasing whenf'(x) > 0, and decreasing whenf'(x) < 0.
PROOF. Obvious from (10) above.8

This result implies that optimal price increases as long as additional sales
increase future demand, and vice versa. This is independent of the cost
function (as a result of zero discount rate assumption). In other words, for a
7See Jeuland (1979), Kalish (1980) and Jeuland (1981b) for more detailed discussions of the
limitation of these types of models for durable goods.
8The theorem is still valid for the more general demand, where there are exogenous shifts:
x = f(x) g(p) h(t) (see Appendix).

144

SHLOMO KALISH

period of positive effects of sales on demand, price is initially low to stimulate

early adopters, which in turn will stimulate demand. Price will monotonically
increase to the point where this 'word-of-mouth' effect diminishes. On the
other hand, if there is a negative effect of sales now on subsequent demand,
price is initially relatively high, skimming more profits from those who are
willing to pay for early adoption, decreasing monotonically over time.
This result is a generalization of Dolan and Jeuland (1981) and Jeuland and
Dolan (1982). They have investigated a situation with particular functional
form for f(x) and g(p). Here we show that these results apply to any model
regardless of the particular specification, as long as it is separable, i.e. it is not
a peculiarity of the functional form they have chosen.
A typical situation previously examined is the case of a durable good where
at introduction positive word of mouth stimulates demand, while later in the
life cycle, saturation takes over and demand increases. In this case optimal
price under the above conditions will start relatively low, increase as long as
the word-of-mouth effect overcomes the saturation effect, and decrease thereafter. The case where word of mouth is not strong enough to overcome the
saturation effect, or when it has an adverse effect (bad product), price will
monotonically decline. Finally, for a nondurable product, where there is no
saturation effect, price will start relatively low, and monotonically increase as
the product gains reputation.
As for the sign of the shadow price, X, note that in this case we can integrate
(9b) to get:
X(t) = c(t)-

dr

c(T)-

which simply states that marginal benefit for producing more units now is the
cost savings plus the benefits due to effects on demand. Thus in the third case
above, wheref'(x) > 0 uniformly, clearly X(t) is positive and decreasing, so the
difference between marginal cost and marginal revenues is big initially, and
monotonically decreases to 0. The opposite case where f'(x) < 0 uniformly,
there is a tradeoff between future cost savings and adverse effects on demand.
However if there is no cost decline, the shadow price will be negative and
increasing to 0, which now implies pricing higher than the marginal revenues
equal marginal cost rule.
EXAMPLE 2.

Consider again g(p) =

c+-f'

=-c(x(T))

e-dp.

(9b) turns out to be:

lf =f_C+
____?
T+
d )X(t)

(-d)

+ c(x(t)) + 1 ln(f(x(T)))
d ln(f(x(t)))

dt
dt

145

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

s
c(T)

+ 1/d
.---^.PRICE
*
T
TIME

I
0
a. Weak Word of Mouth
(a > bN)

I
0

,
E
T ME

b. Strong Word of Mouth

Long Horizon
(a < bN)

I
T

TIME

c. Strong Word of Mouth

Short Horizon
(a < bN)

FIGURE 2. Optimal Price for x = [N - x][a + bx] e-d. Note that c(T) will be generally
different across these situations. Therefore, these figures should not be compared quantitatively.

So

=c-X+

x(T
c(x(T))

ln(f(x(T))) d
ln(f(x(T)))

In the case f(x) = (N - x)(a + bx), which Dolan and Jeuland (1981) have
investigated, this result reduces to their result. It is easy to verify that in this
case f'(x) > 0 as long as x/N < (1 - a/bN). Thus, if a < bN,9 increasing
price is optimal until the above point, and then price should decrease (see
Figures 2(b) and (c)). If, however, a > bN, then f'(x) < 0 Vt, which implies
price decreases monotonically (see Figure 2(a)).
3.2.2. Separable Demand, Positive Discount. The more realistic case of
r > 0 complicates the analysis. Clearly, the higher the interest rate, the more
valuable early profits are in comparison with future profits. Therefore price
will be shifted from the nondiscount price towards the myopic monopolist's
price, trading more profits now for less profits later. At the extreme, if r is very
large, price will be close to that of a myopic monopolist. As we have shown
earlier in the absence of demand-experience interaction, this causes prices to
follow cost, i.e. decrease over time. However, when experience interacts with
demand, the resultant price depends on the tradeoff between the two factors.
For cases where the two factors reinforce each other, e.g. negative word of
mouth, the resultant price will indeed decrease over time as we will show. For
the more interesting case of positive word of mouth, however, the resultant
price path depends on the tradeoff between the two factors.
To unbundle these interactions, we shall examine first the situation where
there are no learning economies, and then combine the two. (In order to
improve clarity, only several 'typical' situations are presented here, while the
proofs and comments on other cases are deferred to the Appendix.)
THEOREM3.

For the separable demand, with constant cost over time, if

0
E
<
Vt
[0, T], then optimal price is characterizedas follows:
f"(x)
(a) If f'(x(T))> 0=p > 0 VdtE [0, T].
9In empirical studies this condition usually holds. See Bass (1969), (1980).

146

SHLOMO KALISH

(b) If f'(xo) > O,f'(x(T)) < 0=p is increasing and then decreasing.
(c) If f'(x) < 0==p is monotonically decreasing, or increasing and then
decreasing.
PROOF. See Appendix.

Note that the inclusion of interest rates still keeps the qualitative results
similar to the no interest case. Consider again the 'Bass'-type diffusion effect,
f(x) = [N - x][a + bx]; f"(x)= -2b < 0 here. So if the planning horizon is
such that f'(x) > 0 for the whole period, price will monotonically increase. If
the planning horizon is longer, price will increase and then decrease. However,
if the word of mouth is weak to start with, i.e. bN < a, price will either
monotonically decrease, or increase and then decrease.
Finally, for the case where we add experience, we can characterize the
following situation:
THEOREM 4. For the separable demand with learning cost decline and positive
discount, if f"(x) > 0, the following characterize optimalprice:
(a) Iff'(x(T)) < Op < 0 Vt E [0, T].
(b) If f'(x(O)) < 0 and f'(x(T)) > 0 =p* is decreasing and then increasing.
(c) Iff'(x(O)) > 0 = price is either monotonicallyincreasingor decreasing and
then increasing.

PROOF. See Appendix.

Note that the results given in Theorem 4 above are for the less interesting
case, since in many situations we expect f"(x) < 0 (e.g. the Bass model). In
what follows we will discuss the other case and summarize the separable
demand function.
3.2.3. Summary of Separable Demand Pricing. By examining several special cases we have been able to isolate how each of the factors-cost decline,
demand diffusion-saturation, and discount rate-affect price over time. We
have shown that if discount rate is negligible, then price increases as long as it
is beneficial to 'subsidize' adopters, i.e. demand increases due to more units
sold and vice versa. This policy makes intuitive sense, since it will also tend to
smooth production: increase price in anticipation of increasing demand, thus
mitigating the sales increase and vice versa.
Since discount rate means that money now is preferred to future income,
this policy is shifted slightly towards that of a myopic monopolist. At the
extreme, if r is very large, price would be practically identical to that of the
myopic monopolist, because mostly immediate profits would count. Such a
myopic policy follows the well-known marginal revenues equal marginal cost
rule. Since cost declines with production, price would fall along with cost.
Thus, it is the interaction of these two factors that will determine price in the
realistic setting. If the two factors reinforce each other, when demand is
decreasing with more sales for example (which is the case of durable goods
later in the life cycle), then price decline is called for by the two factors: cost
and demand. If, on the other hand, word of mouth has a strong effect in the
introduction of such a durable, then low penetration price is possible, then
increasing, followed by the decreasing price later on. Finally, if the demand is

147

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

Demand
Interaction

Cost and
Interest

r>
c (x)

r =O. Any Cost

(No Discounting)
n

r>
c (x)<0

(No Cost
Decline)

1. Positive Wordof-Mouth Throughout.

Undetermined
2

f(x)

01

P*

I
T

IIT

(If f (x)-0)

2. Positive Wordof-Mouth Followed by Saturation.

Undetermined
If r c (x)
very big

"
f(x)

p/^^
I
x

(If f (x) O)

I.
If 'strong
T
Word-of Mouth

3. Negative Effects
of Penetration
on Demand.
A

f(x)

P
,
I *
T

(If f (x)< O)

.' '
I
T

(If f

'

(x) -O)

I?
I
T

FIGURE3. Summary of Pricing over Time: Separable Demand Function, under Different
Cost/Demand Experience Interactions.

increasing with more sales, price change over time depends on the particular
case: how high the interest rate is, the rate of the cost decline, and the
intensity of the 'word-of-mouth' factor. In case there is no cost decline, e.g.,
price indeed will monotonically increase (see Figure 3).
Finally, consider the shadow prices. If the 'word-of-mouth' effect is positive
throughout, then both cost decline and word of mouth creates an incentive to
'invest' in over production. Indeed, the shadow price is positive, which means
that price is lower than the myopic monopolist's price facing the same
conditions, or in other words, price is below the price that will maximize
immediate revenues at each point in time. Moreover, the discounted value of
the shadow price is monotonically decreasing to 0 at the end of the planning
horizon.
On the other extreme, if each additional unit sold has an adverse effect on
demand, durable goods e.g., then there is a conflict between the cost benefit
on the one hand, and the demand decline on the other, so one cannot tell in
general whether price will be above or below the profit maximizing price.
However, if there is not cost decline, price will always be above the immediate
profit maximizing price.

148

3.3.

SHLOMO KALISH

Other Functional Forms

As pointed out earlier, the separable demand function has several limitations. In particular, it is limited for modeling durable goods. In this section we
will analyze two functional forms that seem more appropriate in the context of
durable goods.
For a perfectly durable good (i.e., goods that last forever), the market size is
limited by the number of individuals who are willing to purchase the durable
at a particular price. Therefore, it is logical to characterize N(p) as the
number of units sold over the life cycle of the product, i.e. the total number of
individuals who are willing to buy at a certain price. Sales rates or the timing
of adoption, however, is not instantaneous. In general, the sales rate is a
function of the remaining potential at a given price, multiplied by the
likelihood of purchase. In the case where the product is new, the likelihood of
purchase increases as a function of cumulative sales since information about
the product is spread by word of mouth and the product self-advertising,
thereby reducing the risk associated with buying the product.'?
Mathematically we have x = [N(p) - x] h(x), where [N(p)- x] is the
remaining potential, while h(x) is the conditional likelihood of purchase." For
a good product we expect h'(x) > 0, i.e. the conditional likelihood of purchase
increases with more products sold.
These models have an appealing interpretation as follows. Each individual
in the population is characterized by a reservation price for the product, which
is the highest price he would be willing to pay for the product. Given an actual
price p, N(p) represents the total number of users whose reservation price is
above p. As price decreases over time, more and more people find price
acceptable. However if price increases, some who have already bought may
find out that the selling price p is above their valuation of the product. In this
case, they may find it beneficial to sell the durable on the secondary market.'2
Therefore the relevant market potential for the firm is still [N(p)- x]. At the
extreme case, if price is increased so that N(p(t)) < x(t), then sales halt
(unless they are willing to buy back at this price, e.g., DeBeers has recently
bought back diamonds in order to keep prices high (Fortune (1982))). The
model can be modified as follows:
=([N(p)-x]
0

h(x)

for N(p)>x,
otherwise.

However for simplicity, we will proceed by assuming that the optimal price
'OProduct self-advertising means that the product itself by being out there (e.g. a new car)
generates awareness. See Kalish (1980) for an analysis of such a model, where word-of-mouth
effect is modeled as [a + bx].
"Note that this form cannot be written as f(x) *g(p), therefore it is not multiplicatively
separable. In fact, the elasticity of demand here crucially depends on the level of penetration, x.
See Kalish (1980) for more details.
12As an example, consider the reselling of silver and gold by individuals when gold and silver
prices skyrocketed in 1979.

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

149

will never be increased to the point where sales actually halt, in which case we
can proceed without this modification.13
The case where r = 0 represents perfect price discrimination. In other
words, the firm will price initially at the highest reservation price. Once these
are sold, it reduces the price to the second-highest reservation price, and so on
until the point where price equals marginal cost. Thus, regardless of h(x), if
there is no time preference for money, the price is monotonically decreasing
with perfect price discrimination, i.e., there is no consumer surplus.
The situation is different for r > 0 when the firm prefers more cash flow
early on. Therefore, it has the incentive to reduce price at the beginning. In
what follows, we will analyze this situation and find out whether there are
conditions under which penetration pricing is still optimal. We start by
analyzing the simpler case where h(x)= a, and then we look at the more
complicated case.
3.3.1. Simple Price-Timing Modelfor Durables. Here we assume that the
sales rate is proportional to the remaining potential population, i.e. x
= [N(p)- x]a. In this case the system of necessary conditions is:
(a) x = [N(p)-

(b) X= r+

x]a;

x(c(x)-

(c)p*=(c-

x(0)= x0,

X(T)=0,

);

N-

(11)

x)

and the price derivative is:

2-

(N - x)N"
N'2

-rX+2

(N - x)a
'N

(12)

from which it can be verified that p < 0 for r = 0. For r > 0 we can characterize the optimal solution as follows:
THEOREM 5. For x = [N(p) - x]a, if c'(x) < O, then the optimal pricing
policy is monotonicallydecreasing over time.
PROOF. See Appendix.
So as expected, the introduction of positive interest rates does not change
the monotonicity of price decline.
3.3.2. Durable Goods with Wordof Mouth. As we described earlier, in this
case we assume that prior cumulative sales influences the rate of adoption for
13See Jeuland (1981b) for a detailed discussion of related issues, and one way of modeling it.
See also Kalish (1983) for a somewhat different approach.

150

SHLOMO KALISH

those who are potential buyers. The formulation is:

(a) x = (N(p)(b) X = rX+ )

Ii
c'(x) -NI
N'(p)

"I2

(N-

x)- N"

p[~2H (2

(N x) h
+ (N-x).h'\
NH
)h

(N- x)
N'
N'

(c)p=c-X[-

x(O) = x,

x) h(x);

and

(N-x
M=_rA+2x[rh)+2(14)
'[

(13)

(N-

x) h' 1
.
2?
]

< 4)

The case of negative word of mouth is the easier one to characterize as

follows:
THEOREM6. For demand x = [N(p)N
x] h(x), if h'(x) < 0, h"(x) > 0,
and c'(x) < 0 Vx, then the optimalprice is monotonicallydecreasing.
PROOF. See Appendix.

Note that the above sufficient conditions are not binding, i.e., price will
monotonically decrease even under weaker conditions; however these are not
easily interpretable.
The more interesting case, where h'(x) > 0, which is more likely for good
products, offers no uniform answers. In this case, price path would depend on
the relative intensity of the 'word-of-mouth' effect vs. the 'price discrimination' effect, cost decline, and interest rate. If the 'word-of-mouth' effect is
strong at introduction, however, then increasing introductory price is still
optimal. Since it is difficult to characterize the general conditions for which
this will be true, we will illustrate it by an example.
EXAMPLE3. Consider a conditional probability of purchase, given by
h(x) = a + b x. (Such conditional likelihood of purchase was used by
Easingwood et al. (1982).) For simplicity, consider the no interest case (r = 0),
with x(0) = 0. Equation (14) becomes:
(N 2-x)N
p.
P'[ N NN',2
For a < 1, limx_,Ox-

2-x
NI

= oo. Since lim,0(N-

(N- x)bax2(a + bxa)

(15)

x) > 0 (otherwise there would

be no sales), it follows (assuming continuity of price) that p(0) > 0 for some
interval [0, t].
Finally, we will show that obtaining an increasing price under this formulation is likely as compared with the separable demand case. Rewrite (14) as

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

151

follows: (for r = 0)
2/)

(N - x)N"1
/
NV'

= (N - ?/x) * ^'-h4
af
ax
-N'

(16)
(16)

Recall that in the separable demand situation under zero interest price was
increasing iff af/ax > 0. Here we see that the condition for price increase is
stronger, i.e. af/ax > h. This underlines the importance of using the appropriate functional forms in the different cases, as the policy implications could be
substantially different.
With positive interest rates, price shifts towards the short-term profit maximizing policy. This policy is characterized by:
_N-x

N'

_ ( )N - x)N"
N 2N

Note that the price derivative here is monotonically decreasing. Moreover,

even if there is no cost decline, or even cost increase with production, price
still decreases over time, as long as c'(x) < - I/N' Vt E [0, T]. Contrast this
again with the myopic price for the separable demand, where p[2 - gg"/g'2]
= c'(x). x. Here we see that price increases or decreases with costs. So at the
other extreme, where interest rates are very large, we see that with the durable
model there is a stronger tendency for a decreasing price over time.
4.

Conclusions

In their review of pricing models, Monroe and Della Bitta (1978) question
the usefulness of classical pricing models for managerial applications. They
state that there are no models for new product pricing decisions or mature
products.
This work is an attempt to look at the intertemporal pricing implications
due to what we consider to be the most important dynamic factors, i.e., word
of mouth for new products, production cost decline with experience, and
saturation (for durable goods in particular). We now state the assumptions
and summarize the results and then discuss the limitations of these assumptions and directions for future research.
4.1.

Summary of Assumptionsand Results

There are three major assumptions made in our analysis. First, we consider
the case of a monopoly that does not concern itself with future entry in its
pricing policy. This assumption could be realistic in situations where there is
patent protection or other high barriers to entry.
The second assumption is that consumers do not anticipate future prices or
have expectations about them, thus altering their behavior. Rather, we assume
that they only react to current price.

152

SHLOMO KALISH

The third assumption is that production cost can be accurately represented

as a function of experience or cumulative production.14
These assumptions allow us to set the problem as follows: the monopolist
wants to determine the unit price over the planning horizon in order to
maximize discounted profits, given that unit cost declines with cumulative
production (learning curve), and that sales rate depends not only on price, but
also on cumulative sales (diffusion and saturation effects).
There are two questions we have focused on: How would this optimal price
compare with the price that will maximize immediate profits at the particular
time period? Secondly, how will the price trajectory over time look?
The price that will be charged by a monopolist that maximizes immediate
profits is the price that equates marginal revenues with marginal cost. We
have shown in the characterization of the general solution that this price is
somewhat altered in our case. The rule here is that immediate marginal
revenues plus the net future marginal benefit should be equal to marginal cost.
The future marginal benefit of an additional unit now is its effect on reducing
future costs, as well as its effect on demand. Thus, in the case where there are
only production cost declines, but static demand, future benefits are always
positive, and therefore price will be lower than the price that will maximize
immediate revenues, except for the last time period.
Similarly, on the sales side, if additional units sold never have adverse effect
on future demand, which is the case for 'good' repeat-purchase goods, then
this factor alone would cause net future benefits to be positive, and price to be
below the profit maximizing price. Clearly, combining cost learning and
'positive' diffusion enhances this effect.
If, on the other hand, there are adverse effects of current sales on future
sales, a durable good for example, then the future marginal benefit of selling
one more unit is negative. Therefore, price will always be above the immediate
profit maximizing price, except for the last time period. However, combining
that with learning curve cost decline and/or with early positive word of
mouth, then the interaction between these factors will determine the net future
marginal benefit. The results cannot be generalized beyond this.
The second question, of price trajectory over time, is difficult to analyze for
the general case. In order to understand the effects of the different factors on
the price trajectory, we analyzed them one at a time. First, for the simple case
where there is only learning on the production side, then for positive discount
rate, price will monotonically decrease over time. Although price is below the
profit maximizing price for each period, price still decreases over time.
A second special case is where interest rates equal zero. In this case, the
whole planning horizon is like one period, since there is no preference for
early profits. We have shown that under this situation, price will be increasing
or decreasing depending on the demand function alone. For a wide variety of
cases, if there is an adverse reaction of current sales on future sales, then price
will monotonically decrease over time. On the other hand, if there is a positive
14Whilethe above assumptions are typical in the literature, the applicability of the results under
different realistic situations will be discussed later.

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

153

diffusion effect for the whole period, then for the subclass of multiplicatively
separable demand functions, price is increasing. Moreover, in such situations,
price is increasing whenever demand is positively affected by additional units
sold and vice versa. Thus, for the typical durable case, where a positive
word-of-mouth effect early in the life cycle is replaced by an adverse saturation effect later on, price will increase initially, representing penetration
pricing followed by decreasing price later on. We have also shown that such
penetration pricing is possible under a different formulation, if initial word-ofmouth effect is strong enough.
Combining the several factors to represent a more realistic situation naturally does not provide clear-cut answers. The higher the interest rate, the
closer price is to myopic profit maximizing price. This, combined with
declining costs, calls for decreasing price over time. If the demand is adversely
affected by more units sold, then these two tendencies reinforce each other,
and in most cases studied here, price is monotonically decreasing over time. If,
on the other hand, there is positive diffusion effect, then the two factors
counteract. However, we have shown that for the constant unit cost situation,
and separable demand functions, the qualitative results remain the same,'5 i.e.,
if there are positive diffusion effects over the whole period, then price
increases over time. If initial positive effects are replaced by negative effects
later on, then price increases and then decreases. For the general case, not
much can be said.
4.2. Applicabilityand Validity
Any model is a simplified representation of the real world. A successful
model, however, must include the most important and relevant factors.
Clearly, there is no situation where our simplifying assumptions hold absolutely. However, in cases where this is a good approximation, the model's
implication should be applicable. (This, of course, can be tested with actual
data, as will be outlined below.)
First, consider the no competition assumption. We have assumed here that
demand depends on the monopolist price alone. Clearly, in the case of
competition, demand will depend on competitors' price as well. In some
situations, patent protection as an example, the monopoly assumption may
serve as a good approximation. If there are no such barriers to entry, then we
have to consider the effects of these recommendations on the firm's competitive position. For example, we saw that learning curve cost decline causes the
firm to reduce price, thus investing in higher return in the future, as cost will
be lower. Similar results are obtained on the demand side where positive
word-of-mouth effect can be regarded as reputation. So, in these cases, the
monopoly model recommendation improves the firm's competitive position
relative to a myopic pricing policy by creating better market reputation and a
cost advantage. (This is conditional on the appropriability of these investments, i.e., to what extent there are spill-over effects in cost production and
product reputation.)
15See Theorem 3 for exact details and additional restrictions.

154

SHLOMO KALISH

Consider, next, the consumer expectation problem. To what extent are

people going to change their behavior by anticipating future prices? This
could be a central issue in cases where prices change rapidly. If price is
decreasing over time, many people delay purchases (e.g. calculators, microcomputers), while if price is increasing, many individuals make early purchases (e.g. it is a common phenomenon in countries with a high rate of
inflation, where prices are increasing frequently; usually there are long lines of
buyers prior to such increases). However if the resulting optimal price does not
change dramatically over time, this assumption can still be a good approximation for the real world.
The third assumption, that unit cost depends on cumulative output alone, is
less restrictive in my judgment, although this has to be verified for each
particular case.
Do the results of this study provide insight to planners for pricing products
over time? Suppose the three assumptions above hold, then there are several
policy implications that stand out:
1. Learning cost and positive diffusion effect cause prices to be lower than
the immediate profit maximizing price.
2. Learning cost alone, with no sales dynamics, causes price decline over
time.
3. Positive diffusion effects for the whole period, with constant cost, cause
low introductory price which increases later on (see Theorem 3 for exact
details).
4. Durable goods price tends to fall over time, thus price discriminating over
time. If initial word-of-mouth effect is strong enough, a lower introductory
price, increasing and then decreasing, is possible.
5. If the separable demand function is a good approximation, and interest
rate is low, then the policy guideline is as follows: Price is increasing as long as
there are positive effects of sales on future demand, and vice versa.
This policy guideline is also appealing since it stabilizes production cycles,
i.e., price is low if demand is lower and increasing, price increases as long as
demand is increasing, and vice versa. Thus, price increases in anticipation of
high demand, reducing production, and vice versa.
These conclusions are consistent with and unify the current state of knowledge. In particular, consider the two different types of models that were
recently analyzed by Bass and Bultez (1982), and Dolan and Jeuland (1982)
respectively. Bass and Bultez postulate that the diffusion effect is a function of
time since introduction, while in the other formulation it is a function of the
number of adoptors. Here we show that the results for particular parameterization reported in these two papers are actually true for general classes of
problems; i.e., if changes in demand over time are a function of time alone
then price would decrease monotonically given production cost decline. If,
however, the expansion in demand is a result of the number of adopters, then
it may actually be optimal to price low and increasing at introduction. We
have also shown that for durable goods increasing price at introduction is not
as likely as might be perceived by using the separable model, although it is a
possibility if the "diffusion" effect is strong enough.

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

155

As for the bottom line, does it make a big difference in term of the present
value of profits? Robinson and Lakhani (1975) report substantial differences
between a myopic policy and the optimal policy, while Bass and Bultez (1982)
report small differences. We can provide some insight into understanding
under what conditions there would be big differences. Recall that X(t), the
shadow price, measures the marginal effect of the number of adopters at time
t on the objective function, the discounted profits. While a full analysis of the
parameters that affect the sensitivity of deviation from optimal policy on
profits is not presented here, the following are some rules of thumb: (1) The
higher the discount rate, the smaller the difference between myopic policies
and optimal policies. This is intuitive, since if future profits are heavily
discounted, then most that counts are early immediate profits. (2) The higher
the rate of cost decline, particularly at introduction, the larger the difference
in the bottom line. Again, this is intuitive, since if cost decline is large, then
every unit produced affects future profits by lowering future costs. (3) The
larger the positive diffusion effect is, the larger the difference between the two
policies. Again, this is logical similarly to the above. Therefore, it is not all that
surprising that Bass and Bultez report small differences, since they have used
r = 0.30, which is a large discount rate (note that the relevant interest rate here
is the real interest rate, after subtracting the inflation rate). Also in their
formulation there are no diffusion effects, since the shifts in demand are a
function of time, and not of the number of adopters. While the discount rate
in Robinson and Lakhani's example was high as well (0.40), and the cost
decline was similar, their formulation has the diffusion effect which made the
difference.
In any real situation there are other marketing variables at the firm's
disposal. Advertising can substitute for word of mouth, warranties can substitute for reducing uncertainties, etc. Therefore stimulating early adopters need
not be done by price alone. However, since these factors exhibit decreasing
return to scale, it is the author's conjecture that even in their presence the
qualitative results presented above will still hold. That is, even if advertising is
set at the optimal level, it is still more efficient to somewhat subsidize early
adopters if their effect on increasing future demand is positive. This conjecture
has support in Kalish (1983). It is also the author's conjecture that violation of
the competition assumption, and the consumer expectations assumption will
not change the qualitative nature of the results, but this is left for future work.
Finally, do we observe these pricing patterns in practice? Although a
full-scale empirical analysis remains to be done, casual observations are
consistent with the above implications:
*We observe price decline in situations where cost decline is a major factor
(calculators, semiconductors).
*We observe low introductory prices for repeat purchase goods (coupons,
free samples, specials).
*In durable goods, where 'penetration' pricing is less likely, we observe
mainly price decline over time (televisions, Rubik's cube, cameras).
* In cases where initial word-of-mouth is crucial, we observe low introductory price (medical instrumentation, e.g., cat scanners).

156

4.3.

SHLOMO KALISH

Future Research

The main direction for future research is the introduction of competition.

This is a very difficult task, due to the complex interaction between competitors over time, and uncertainty about future entry. However, several attempts
have already been made (see Teng and Thompson (1980), Eliashberg and
Jeuland (1982), Clarke and Dolan (1982), Spence (1981), and Bass and Rao
(1982)).
A second direction is relaxing the assumption that cost declines with
cumulative production alone. A recent attempt in this direction has been
published recently (Clarke et al. (1982)). The 'nonanticipating consumer'
assumption can be relaxed as well. There are several ways to introduce
consumers' expectations (see, e.g., Bulow (1981), Schmalensee (1979)). Finally,
an important direction for future research is the inclusion of other marketing
mix variables, e.g. advertising. (Some recent attempts are Horsky and Simon
(1982), Kalish (1983), Spremann (1981), and Teng and Thompson (1980).) In
this context, heterogeneous population with respect to advertising and socialcommunication can be explored (see Jeuland (1981a)).
Appendix. Optimal Solution Characterizationand Proof of Theorems
We assume f(x, p) and c(x) to be twice continuously differentiable, fp < 0,
0 Vp E (- oo, oo). We assume that a solution exists. Since the problem is
>
f
unconstrained, the necessary condition at the optimal price is Hp = 0. (A
general existence proof is difficult for this unconstrained problem. However, it
is natural that any well-posed realistic problem has a solution indeed.)
Moreover, at the optimal path, a necessary condition for a maximum is
Hp < O, therefore it follows that (5) is true. We will use (5) frequently in the
proofs.
The solution is not necessarily continuous, let alone differentiable. However, we assume the solution is twice differentiable in some of the proofs. For
cases where the Hamiltonian is concave, however, then the necessary conditions above uniquely define the optimal solution, which ensures continuity
and differentiability of the optimal solution. This condition can be checked
for particular cases. It is conjectured that for most well-posed realistic problems the solution will be continuous and 'differentiable'.
Consider the more general case where we allow for exogenous shifts in
demand over time, x = h(t) . f(x, p). Characterization of optimal solution and
price derivative for this system is similar to (2):
(a) x = h(t) . f(x, p);

(b)A= rA+( fx +c)

(c)p =c-A-

fp

JP

(d)=
4
L

x(0)=

x X;

xo,

r 2

2
J

(A1)

(T)=0,

L fpfP

42p

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

157

In the proofs that follow, we shall see that Theorems 1 and 2 hold for the
more general system above.
Set all f-- 0 in (Al) above.
I. Since c'(x) < 0, >X(t) > 0 Vt E [0, t). Therefore, by (Al)(d)p < 0.
II. Since X(t) > 0 Vt E (0, T), for a given x, the net present value maximizer
(NPV) will price lower than the myopic monopolist (MYP). Therefore, at
PROOF OF THEOREM 1*.

t = 0, XNPV > XMYP. Suppose by contradiction, 3t s.t. XNPv(t) = xMyp(t), then

< 'XMYP _ , a contradiction. Since for any given
B3x < x < x(t) s.t. - NPV
that average NPV price is lower than average
it
is
clear
<
X, PNPV(X) PMYP(X),

MYP price.
III. e- r \X(t)= ftc'(x)

* Xdr, which is clearly decreasing.

(Al)(d) becomes:

PROOFOF THEOREM2*.

P2[

PROOF OF THEOREM 3.

=-

From

F-

g .h(t).

(10) it is obvious

that sign(/3(T))=

sign(f'(x(T))). Taking the time derivative of (10), and substituting p = 0,

c = 0, we get that the second derivative of the optimal path at such point is:

/2-^

g2

1=-^-^
=
.:. g,

If f"(x) < 0, then:

(a) f'(x( T)) > O==p(T) > O. Assuming twice differentiability of p(t), if p(t)
< 0, contradiction. =p=(t) > 0
changes signs, p < 0 at that point, =p(T)
Vt E [0, T].
(b)p(T) < 0. Let tbe the point that f'(x(0) = 0.

(t )

e-r

(-gt

dt T< o=

p(t

) >o

by (10). From (a), p(t) > 0 Vt E [0, t]. Assuming regularity,p(t) cannot change
signs more than once after that, which completes the proof.
(c) Here p(T) < 0, but one change of sign cannot be ruled out.
Note also that if f"(x) > 0 Vx, then by a similar argument:
(a) If f'(x(T))

< O=p(t)

< 0 Vt E [0, T].

(b) If f'(x(O))< 0 and f'(x(T)) > 0, then p(t) is decreasing and then
increasing.
(c) If f'(x(O)) > 0, then p(t) is either monotonic increasing, or decreasing
and then increasing. CI

158

SHLOMO KALISH

In case we get that the second derivative of price at

PROOF OF THEOREM 4.

= 0 is:

I" -g" gg
4[2
oZ-

--

. g2 x -r
f, r,

g,2

The theorem follows now as 3 above.

PROOFOF THEOREM5.
PROOF OF THEOREM6.

c'(x). x.

g'

This is a special case of Theorem 6.

Substitute X from (13c) into (14), taking time

derivative, and substituting 3 = 0, we get:

P. 2

(N-x)N"
N,2

- [r. (c(x) +

) + N(2h-4h'(N-x)

+ (N-x)2

h).

Since all terms in brackets are negative, jl 0 > 0. From (14), p(T) < 0, so p(t)
cannot change signs under the regularity conditions. O
References
Abell, D. F. and J. S. Hammond (1979), Strategic Market Planning, Englewood Cliffs, N.J.:
Prentice-Hall.
Alchian, A. (1959), "Costs and Outputs," in The Allocation of Economic Resources, M. Abramovitz et al., eds., Stanford, Cal.: Stanford University Press, 23-40.
Arrow, K. J. (1962), "The Economic Implications of Learning by Doing," Review of Economic
Studies, 29 (June), 155-173.
and M. Kurz (1970), Public Investment, the Rate of Return, and Optimal Fiscal Policy,
Baltimore and London: The Johns Hopkins Press.
Bass, F. M. (1969), "A New Product Growth Model for Consumer Durables," Management
Science, 15 (January), 215-227.
(1980), "The Relationship Between Diffusion Rates, Experience Curves, and Demand
Elasticities for Consumer Durable Technological Innovations," Journal of Business, 53 (July),
S51-S67.
and A. V. Bultez (1982), "A Note on Optimal Strategic Pricing of Technological
Innovations," Marketing Science, 1 (Fall), 371-378.
and Ram C. Rao (1982), "Equilibrium Dynamic Pricing of New Products in Oligopolies:
Theories and Evidence," Working Paper, University of Chicago, Graduate School of Business.
Bulow, J. I. (1981), "Durable Goods Monopolists," Working Paper, Stanford University, Graduate School of Business (March).
Clarke, D. G. and R. J. Dolan (1982), "A Simulation Model for the Evaluation of Pricing
Strategies in a Dynamic Environment," Working Paper, Harvard University, Graduate
School of Business Administration.
Clarke, F. H., M. N. Darrough and J. Heineke (1982), "Optimal Pricing Policy in the Presence of
Experience Effects," Journal of Business, 55, 4 (November), 517-530.

MONOPOLIST PRICING: DYNAMIC DEMAND AND PRODUCTION COST

159

Conley, P. (1981), "Experience Curves as a Planning Tool," in Corporate Strategy and Product
Innovation, R. Rothberg, ed., 2nd Ed., New York: The Free Press, 132-143.
Dolan, R. J. and A. P. Jeuland (1981), "Experience Curves and Dynamic Demand Models:
Implications for Optimal Pricing Strategies," Journal of Marketing, 45 (Winter), 52-73.
Easingwood, C. J., V. Mahajan and E. Muller (1982), "A Non-Uniform Influence Innovation
Diffusion Model of New Product Acceptance," Working Paper (August).
Eliashberg, J. and A. P. Jeuland (1982), "New Product Pricing Strategies Over Time in a
Developing Market: How Does Entry Affect Prices," Working Paper, University of Chicago,
Graduate School of Business.
Fortune (1982), "DeBeers and the Diamond Debate," (September 6), 42-53.
Haley, M. J. (1981), "A Monopolist that Learns by Doing," Working Paper, Vanderbilt University, The Owen Graduate School of Management (September).
Henderson, B. (1980), Henderson on CorporateStrategy, Boston: Abt.
Hirsch, W. (1952), "Manufacturing Progress Functions," Review of Economics and Statistics, 34
(May), 143-155.
Horsky, D. and L. S. Simon (1983), "Advertising and the Diffusion of New Products," Marketing
Science, 2, 1 (Winter), 1-17.
Jeuland, A. P. (1979), "Epidemiological Modeling of Diffusion of Innovation: Evaluation and
Future Directions of Research," in 1979 Educators' Conference Proceedings, N. Beckwith et
al., eds., Chicago: American Marketing Association, 274-278.
(1981a), "Parsimonious Models of Diffusion of Innovation. Part A: Derivations and
Comparisons," Working Paper, University of Chicago, Graduate School of Business.
(1981b), "Parsimonious Models of Diffusion of Innovation. Part B: Incorporating the
Variable of Price," Working Paper, University of Chicago, Graduate School of Business.
and R. J. Dolan (1982), "An Aspect of New Product Planning: Dynamic Pricing," TIMS
Studies in the Management Sciences, Special Issue on Marketing Planning Models,
A. Zoltners, ed.
Kalish, S. (1980), "Models of the Diffusion of Innovations and their Implications for Government
Policies," Master's Thesis, Massachusetts Institute of Technology (June).
(1983), "New Product Diffusion Model with Price Advertising and Uncertainty," Working
Paper, Graduate School of Management, University of Rochester, MERC83-02 (January).
Monroe, K. and A. Della Bitta (1978), "Models for Pricing Decisions," Journal of Marketing
Research, 15 (August), 413-428.
Robinson, B. and C. Lakhani (1975), "Dynamic Price Models for New-Product Planning,"
Management Science, 21, 1113-1122.
Rogers, E. M. and F. F. Shoemaker (1971), Communication of Innovations: A Cross-Cultural
Approach,2nd Ed., New York: The Free Press.
Rosen, S. (1972), "Learning by Experience as Joint Production," QuarterlyJournal of Economics,
86 (August), 366-382.
Schmalensee, R. (1979), "Market Structure, Durability and Quality: A Selective Survey," Economic Inquiry (January), 177-196.
(1982), "Product Differentiation Advantages of Pioneering Brands," The American Economic Review, Vol. 72, No. 3 (June), 349-365.
Simon, H. (1982), "PRICESTRAT-An Applied Strategic Pricing Model for Non-Durables,"
TIMS Studies in the Management Sciences, Special Issue on Marketing Planning Models,
A. Zoltners, ed. (forthcoming).
Spence, A. M. (1981), "The Learning Curve and Competition," Bell Journal of Economics,Vol. 12,
No. 1, 49-70.
Spremann, K. (1981), "Hybrid Product Life Cycles and the Nerlove-Arrow Theorem," Working
Paper, Abt. Wirtschaftswissenschaften, Universitat Ulm, D-7900 Ulm, Federal Republic of
Germany (July).
Teng, J. and G. L. Thompson (1980), "Optimal Pricing and Advertising Policies for New Product
Oligopoly Models," Working Paper, Carnegie-Mellon University.